The Pentagon has launched a major examination of civilian deaths in military operations, responding to criticism that it has failed to protect innocent bystanders in counterterrorism wars worldwide.

The far-reaching initiative to create the military’s first-ever policy on civilian casualties, which senior Pentagon officials began last year, seeks to answer a central question: Why is the military’s estimate of civilian deaths so much smaller than outside tallies?

Last week, the Pentagon reported that 1,190 civilians had been killed by American strikes in Iraq and Syria since the beginning of the campaign against the Islamic State in 2014. Airwars, a respected monitoring group, put the figure at more than 7,200 dead, more than six times as high.

The effort is underway as the Pentagon races to conclude its campaign against the militant group, unleashing a torrent of airstrikes ahead of President Trump’s ordered withdrawal from Syria. While officials have described the targeting of the ­Islamic State as the most precise in history, a high civilian death toll has fueled questions about whether the president’s bare-knuckled approach has resulted in greater loss of life.

Over the past year, officials from across the military have reviewed the way the Pentagon plans and conducts airstrikes, its procedures for handling allegations of civilian deaths, and decisions about when to acknowledge errant strikes. The assessment, which includes a classified study commissioned by Gen. ­Joseph F. Dunford Jr., chairman of the Joint Chiefs of Staff, comes as lawmakers press the military to improve its handling of noncombatant deaths.

That study, whose existence and findings have not previously been made public, recommends a more open, standardized investigations process, but does not seek to determine the root cause of a spike in casualties during the peak of the operation against the Islamic State.

Watchdog and advocacy groups see the effort as a hopeful sign but remain concerned that it could reaffirm existing problems or fall short of the substantial change Pentagon leaders say they want.

“After two years of watching the death toll grow, it’s really tempting to be satisfied that such a study took place,” said Daniel Mahanty, director of the U.S. Program at the Center for Civilians in Conflict (CIVIC), an advocacy group. “While we’re happy that Pentagon leaders saw this as an important issue, our focus now is to make sure it results in meaningful changes to prevent casualties and ensure those who have been killed get the acknowledgment they deserve.”

The attempt to determine a more accurate picture of the impact of operations on civilians — and to codify steps to prevent deaths in the first place — nonetheless represents a milestone close to two decades after the United States launched its global counterterrorism wars in the wake of the Sept. 11, 2001, attacks.

“This is a massive undertaking, and it’s about freaking time,” said one former official familiar with the initiative, who, like others, spoke on the condition of anonymity to describe internal discussions. “This should have happened in 2002.”

'Opening salvo'

In late 2017, discussions about civilian deaths had reached a fever pitch at the Pentagon’s highest levels.

While the military had developed a system to keep civilians safe, including extensive measures to surveil targets and calculate the damage from explosions, senior officials knew they had a problem.

The battles to liberate the ­Islamic State’s twin capitals — Mosul in Iraq and Raqqa in Syria — included some of the most significant urban operations involving the United States since the Vietnam War.

Unlike its previous war in Iraq, the Pentagon had only a small ground presence. Tiny cells of Americans worked to advise local forces pressing into military territory, presenting an additional challenge for air operations.

Galvanizing a sense of crisis was a massive March 2017 bombing in Mosul that killed civilians. In an indication of the confusion that at times has characterized the U.S. response, officials initially said they were unsure if a U.S. strike was responsible for the destruction. Later, an investigation found that an American bomb had struck a building where militant snipers were positioned, setting off secondary explosions that led to its collapse and killing more than 100 people sheltering inside.

Much of the scrutiny surrounded Trump’s steps to empower commanders after he promised a swift end to the war and suggested the United States should “take out” militants’ families. Then-
Defense Secretary Jim Mattis stressed the Pentagon’s commitment to keeping civilians safe. He also pushed back against critics, saying they were trying to hold the military to a standard of zero errant deaths, something “that has never been achieved before in warfare.”

“I’m never okay with any civilian casualty,” Mattis told reporters in a heated December 2017 exchange. “Don’t screw with me on this.”

Behind the scenes that same month, Mattis took the unprecedented step of convening aid and watchdog groups at the Pentagon to confidentially discuss the plight of noncombatants. In that meeting, Mattis said he wanted to make sure his forces were equipped to minimize civilian harm.

“This was an opening salvo that they were taking this seriously,” one participant said.

'A mixed bag'

In early 2018, Dunford commissioned a study to address the accounting gap. That spring, a team of officials and external experts, led by scholars from the National Defense University, interviewed targeteers and analyzed data from thousands of airstrikes.

“One of the big arguments we were making was, ‘How do you possibly know if you are minimizing civilian harm if you don’t study it?’­ ” the former official said.

While the Pentagon had decided against making the study public, Dunford later ordered it to be partially declassified in response to The Washington Post’s plans to write about it.

Among its more critical findings, the study says the military has not adequately used outside information to verify whether civilians have died. It also found that processes for examining allegations varied between geographic commands.

Recommendations included clarifying “guidance and doctrine” in operations that rely on partner forces; bolstering investigation cells; and developing a system for condolence payments or making amends.

On other issues, the study found existing procedures to be adequate and did not recommend changes to tactics or rules of engagement.

Sarah Margon, Washington director at Human Rights Watch, called the report “a mixed bag.” “It’s significant they’re pursuing a larger policy for the Pentagon, but the gaps in there and the sanitized findings make it questionable that they can really produce something meaningful,” she said.

Individuals familiar with the study described a disagreement between its authors over how critical it should be. Some believed it missed an opportunity to more directly address shortcomings; others said a scathing analysis might lead operational commanders to dismiss it out of hand.

Those differences are visible in discussion of the system for distinguishing between combatants and civilians. Although the study states that the “positive identification” process — which typically relies on drone imagery or intelligence — “has sufficient guidance and structure and therefore does not increase the risk for civilian casualties,” that assertion is disputed in a lengthy footnote by several authors who characterize it as a primary culprit.

If investigators rely on the same information to investigate a strike as they did to rule out the presence of civilians ahead of time, they argued, how could they possibly reach a conclusion that civilians died?

While the study ruled out several factors as being responsible for the increased civilian bloodshed, it stopped short of addressing those that were.

Larry Lewis, an expert on civilian casualties and an author of the study, said he believed the main drivers included a decision late in the Obama administration to increase the noncombatant value, a figure that represents the highest number of civilians strike planners can put at risk without seeking higher approval. He also pointed to what he characterized as “command emphasis,” which he said had resulted in reduced civilian deaths in Afghanistan in the past.

Senior officials said they “took it as a given” that more strikes would result in more accidental deaths. “If you drop 10,000 bombs in a five-square-mile area,” a senior Joint Staff official said, “then you’re going to see a greater effect against the enemy and also some greater impact on civilians.”

A demand for action

After the study concluded last spring, officials began using the recommendations to develop a new policy. That effort, which has included tabletop exercises and workshops on such issues as condolence payments, is expected to conclude in late 2019.

As the process got underway, Congress was moving to require greater action. Responding to the spike in civilian deaths, lawmakers included measures in the annual defense bill to name a Pentagon coordinator and mandate greater transparency. Advocates saw the selection of David Trachtenberg, a higher-ranking official than the legislation required, as a positive move.

In a recent interview, Trachtenberg said the department would consider a number of changes, from procedures for planning air raids to post-strike assessments.

“War is a messy business, so we’re not perfect,” he said. But, “there is a serious commitment to doing what we need to do.”

The Joint Staff has started to update internal manuals, and casualty investigators have increased their use of external information. “We’re not waiting,” the Joint Staff official said.

Officials expressed confidence that the initiative would not be affected by Mattis’s resignation in December. A Pentagon spokesman said Acting Defense Secretary Patrick Shanahan supported the effort “to evaluate and — where possible — improve our ability to minimize civilian harm in our military operations and to be transparent when civilian casualties do occur.”

Although advocacy groups expressed reservations about the scope of the study and policy review, they praised the Pentagon for seeking to improve a system that is far more developed than those of the nations it fights alongside in Iraq and Syria.

But the military has already taken a step back ahead of the withdrawal from Syria, independent monitors said.

The military has reported more than 1,600 strikes in Iraq and Syria since Trump’s Dec. 16 declaration, accounting for 5 percent of American strikes since 2014. But, in a shift, the Pentagon has omitted details on strike dates and locations, making it harder for outside groups to verify casualty reports.

Airwars director Chris Woods described that as a “fairly fundamental” reversal of what has been an increasingly open system. Military officials cited operational concerns but did not provide details.

“Until these issues are embedded in the DNA of the Pentagon and its various military commands, decisions by different parts of the military will be able to impair how the department handles civilian harm,” Woods said.